Effective and efficient gaussian processes
Wednesday 07 Aug 2019
Time: 10:00 - 18:00
Bio
Prior to joining Turing REG, Eric spent 5 years as an Assistant Professor at the Center for Earthquake Research and Information at the University of Memphis. As an academic he taught geophysics courses and led a research group using computational models to study earthquakes.
Eric has a PhD in Computational Physics from the University of California, Santa Barbara, where his thesis research was to perform simulations of earthquakes and material failure. Following his PhD, Eric spent 4.5 years as a researcher at Los Alamos National Laboratory, Université Grenoble-Alpes, and ETH Zurich where he continued to work in geophysics and materials science.
Research interests
Eric's current interests stem from his experience doing computational simulations but finding that they are rarely used in practice to make predictions or forecasts about the real world. This is particularly true in earthquake science, where Probabilistic Seismic Hazard Analysis is based almost entirely on extremely limited empirical data and does not incorporate any knowledge of the physics of earthquakes. In his time as an academic researcher, Eric has spent a lot of time and energy doing advanced computer simulations to understand the mechanics of how earthquakes worked, but there is a significant barrier in going from that information to making practical advances in how scientists estimate the damaging effects of future earthquakes.
At the Turing, Eric led the Research Engineering efforts to tackle this problem by writing software tools that make Uncertainty Quantification methods accessible to domain experts. This is spread over several projects to develop a diverse set of methods and libraries for this approach. Eric's hope is that making these tools part of a robust, well-designed software library will help researchers focus on their specific research goals and help them compute realistic error bars on their simulations to improve confidence in simulations.
